process/scheduler: make interface a little cleaner, don't leak continuation info

src/kernel/entry.c

@@ -140,7 +140,8 @@ static char cmdline_copy[MAX_CMDLINE_LENGTH + 1];
   uint64_t fb_length = ((fb->height * fb->pitch - 1) / 4096 + 1) * 4096;
   fb_physical_base = (uint64_t)fb->address - hhdm_request.response->offset;
   fb_base = find_free_kernel_region(fb_length);
-  map_kernel_region(fb_physical_base, fb_base, fb_length, 1, 0);
+  map_kernel_region(
+    fb_physical_base, fb_base, fb_length, 1, 0);
 
   //store rest of framebuffer information
 
@@ -282,25 +283,12 @@ static const char *cmdline_look_up(const char *key) {
         "calcite/apps/hello/hello.elf") != FAR_SUCCESS)
     panic("could not look up hello.elf")
 
-  struct process *hello = heap_alloc(sizeof(struct process));
-  uint64_t hello_entry;
-  create_process(hello);
-  if (load_elf(hello, &hello_entry, &root_fs, hello_node) != 1)
-    panic("could not load hello.elf")
+  if (!start_elf(&root_fs, hello_node))
+    panic("could not start hello.elf")
 
   if ((*root_fs.free_node)(&root_fs, hello_node) != FAR_SUCCESS)
     panic("could not free hello.elf node")
 
-  struct thread *hello_thread = heap_alloc(sizeof(struct thread));
-  create_thread(hello, hello_thread);
-
-  create_user_task(
-    hello->p4_physical_base,
-    hello_entry,
-    (uint64_t)hello_thread->stack_top);
-
-  running_thread = hello_thread;
-
   resume_next_continuation();
 
 }

src/kernel/process.asm

@@ -0,0 +1,56 @@
+ ; Calcite, src/kernel/process.asm
+ ; Copyright 2025 Benji Dial
+ ;
+ ; This program is free software: you can redistribute it and/or modify
+ ; it under the terms of the GNU General Public License as published by
+ ; the Free Software Foundation, either version 3 of the License, or
+ ; (at your option) any later version.
+ ;
+ ; This program is distributed in the hope that it will be useful, but
+ ; WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ ; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ ; for more details.
+ ;
+ ; You should have received a copy of the GNU General Public License along
+ ; with this program. If not, see <https://www.gnu.org/licenses/>.
+
+
+bits 64
+
+extern running_thread
+
+section .text
+
+;referenced in process.c
+global thread_start
+thread_start:
+
+  mov rax, qword [running_thread]
+  test rax, rax
+  jnz .assert_fail
+
+  mov qword [running_thread], rbx
+
+  mov cr3, rbp
+  mov rcx, r12
+  mov r11, 0x200
+
+  xor rax, rax
+  xor rbx, rbx
+  xor rdx, rdx
+  xor rdi, rdi
+  xor rsi, rsi
+  xor rbp, rbp
+  xor r8, r8
+  xor r9, r9
+  xor r10, r10
+  xor r12, r12
+  xor r13, r13
+  xor r14, r14
+  xor r15, r15
+
+  o64 sysret
+
+.assert_fail:
+  ;could be nice to handle this more gracefully in the future
+  ud2

src/kernel/process.c

@@ -284,6 +284,42 @@ int load_elf(
 
 }
 
+//defined in process.asm. enters user mode with:
+//  running_thread = value of rbx when we jump here
+//  cr3 = value of rbp when we jump here
+//  rsp = value of rsp when we jump here
+//  rip = rcx = value of r12 when we jump here
+//  rflags = r11 = 0x200 (IF)
+//  all other registers zeroed
+extern uint8_t thread_start;
+
+int start_elf(const struct fs_info *fs_info, void *fs_node) {
+
+  struct process *process = heap_alloc(sizeof(struct process));
+  create_process(process);
+
+  uint64_t entry;
+
+  if (!load_elf(process, &entry, fs_info, fs_node)) {
+    destroy_process(process);
+    return 0;
+  }
+
+  struct thread *thread = heap_alloc(sizeof(struct thread));
+  create_thread(process, thread);
+
+  struct continuation_info ci;
+  ci.rip = (uint64_t)&thread_start;
+  ci.rbx = (uint64_t)thread;
+  ci.rbp = thread->process->p4_physical_base;
+  ci.rsp = (uint64_t)thread->stack_top;
+  ci.r12 = entry;
+
+  add_ready_continuation(&ci);
+  return 1;
+
+}
+
 void destroy_process(struct process *process) {
   for (int p3i = 0; p3i < 512; ++p3i)
     if (process->p3_virtual_base[p3i]) {
@@ -307,6 +343,26 @@ void destroy_process(struct process *process) {
   heap_dealloc(process, sizeof(struct process));
 }
 
+void destroy_thread(struct thread *thread) {
+
+  assert(thread->process->n_threads >= 1)
+
+  if (thread->process->n_threads == 1)
+    destroy_process(thread->process);
+
+  else {
+
+    --thread->process->n_threads;
+
+    for (void *p = thread->stack_bottom; p < thread->stack_top; p += 4096)
+      unmap_page_for_process(thread->process, p);
+
+  }
+
+  heap_dealloc(thread, sizeof(struct thread));
+
+}
+
 #define INITIAL_STACK_SIZE (16 << 20)
 
 void create_thread(struct process *process, struct thread *thread_out) {
@@ -319,7 +375,8 @@ void create_thread(struct process *process, struct thread *thread_out) {
   for (int i = 0; i < INITIAL_STACK_SIZE / 4096; ++i) {
 
     uint64_t pma = take_free_physical_page();
-    map_page_for_process(process, pma, stack_bottom_vma + i * 4096, 1, 0);
+    map_page_for_process(
+      process, pma, stack_bottom_vma + i * 4096, 1, 0);
 
     void *kvma = find_free_kernel_region(4096);
     map_in_kernel_page_table(pma, kvma, 1, 0);
@@ -339,28 +396,10 @@ void create_thread(struct process *process, struct thread *thread_out) {
 struct thread *running_thread = 0;
 
 [[noreturn]] void syscall_end_thread() {
-
   assert(running_thread != 0)
-  assert(running_thread->process->n_threads >= 1)
-
-  if (running_thread->process->n_threads == 1)
-    destroy_process(running_thread->process);
-
-  else {
-
-    --running_thread->process->n_threads;
-
-    for (void *p = running_thread->stack_bottom;
-         p < running_thread->stack_top; p += 4096)
-      unmap_page_for_process(running_thread->process, p);
-
-  }
-
-  heap_dealloc(running_thread, sizeof(struct thread));
-
+  destroy_thread(running_thread);
   running_thread = 0;
   resume_next_continuation();
-
 }
 
 void syscall_map_framebuffer(struct framebuffer_info *info_out) {

src/kernel/process.h

@@ -70,7 +70,14 @@ int load_elf(
   struct process *process, uint64_t *entry_out,
   const struct fs_info *fs_info, void *fs_node);
 
+//returns 0 on failure, 1 on success.
+//creates a process and a thread in that process, loads the elf into the process,
+//and schedules a ready task that sets the running thread to the new thread and
+//starts user mode at the elf's entry point.
+int start_elf(const struct fs_info *fs_info, void *fs_node);
+
 void destroy_process(struct process *process);
+void destroy_thread(struct thread *thread);
 
 //returs 1 if [start, start + length) is writable by process, otherwise 0.
 int is_mapped_writable(struct process *process, void *start, uint64_t length);

src/kernel/scheduler.asm

@@ -1,4 +1,4 @@
- ; Calcite, src/kernel/scheduler.c
+ ; Calcite, src/kernel/scheduler.asm
  ; Copyright 2025 Benji Dial
  ;
  ; This program is free software: you can redistribute it and/or modify
@@ -17,30 +17,6 @@
 
 bits 64
 
-;referenced in scheduler.c
-global user_task_start
-user_task_start:
-
-  mov cr3, rbx
-  mov rcx, rbp
-  mov r11, 0x200
-
-  xor rax, rax
-  xor rbx, rbx
-  xor rdx, rdx
-  xor rdi, rdi
-  xor rsi, rsi
-  xor rbp, rbp
-  xor r8, r8
-  xor r9, r9
-  xor r10, r10
-  xor r12, r12
-  xor r13, r13
-  xor r14, r14
-  xor r15, r15
-
-  o64 sysret
-
 ;referenced in scheduler.c
 global resume_continuation
 resume_continuation:

src/kernel/scheduler.c

@@ -19,18 +19,7 @@
 #include "utility.h"
 #include "heap.h"
 
-struct continuation_info {
-  uint64_t rip;
-  uint64_t rbx;
-  uint64_t rbp;
-  uint64_t rsp;
-  uint64_t r12;
-  uint64_t r13;
-  uint64_t r14;
-  uint64_t r15;
-};
-
-static struct continuation_info **ready_continuations = 0;
+static struct continuation_info *ready_continuations = 0;
 static int rc_buffer_length = 0;
 static int rc_read_ptr = 0;
 static int rc_count = 0;
@@ -38,34 +27,30 @@ static int rc_count = 0;
 #define INITIAL_RC_BUFFER_LENGTH 128
 
 void init_scheduler() {
-  ready_continuations = heap_alloc(INITIAL_RC_BUFFER_LENGTH * sizeof(void *));
+  ready_continuations = heap_alloc(INITIAL_RC_BUFFER_LENGTH * sizeof(struct continuation_info));
   rc_buffer_length = INITIAL_RC_BUFFER_LENGTH;
-  for (int i = 0; i < INITIAL_RC_BUFFER_LENGTH; ++i)
-    ready_continuations[i] = 0;
 }
 
-static void queue_continuation(struct continuation_info *info) {
+void add_ready_continuation(struct continuation_info *info) {
 
   if (rc_count == rc_buffer_length) {
 
-    struct continuation_info **new_rc_buffer =
-      heap_alloc(2 * rc_buffer_length * sizeof(void *));
+    struct continuation_info *new_rc_buffer =
+      heap_alloc(2 * rc_buffer_length * sizeof(struct continuation_info));
 
     memcpy(
       new_rc_buffer,
       ready_continuations + rc_read_ptr,
-      (rc_buffer_length - rc_read_ptr) * sizeof(void *));
+      (rc_buffer_length - rc_read_ptr) * sizeof(struct continuation_info));
 
     memcpy(
       new_rc_buffer + rc_buffer_length - rc_read_ptr,
       ready_continuations,
-      rc_read_ptr * sizeof(void *));
+      rc_read_ptr * sizeof(struct continuation_info));
 
-    heap_dealloc(ready_continuations, rc_buffer_length * sizeof(void *));
+    heap_dealloc(ready_continuations, rc_buffer_length * sizeof(struct continuation_info));
 
-    new_rc_buffer[rc_buffer_length] = info;
-    for (int i = rc_buffer_length + 1; i < 2 * rc_buffer_length; ++i)
-      new_rc_buffer[i] = 0;
+    memcpy(&new_rc_buffer[rc_buffer_length], info, sizeof(struct continuation_info));
 
     ready_continuations = new_rc_buffer;
     rc_buffer_length *= 2;
@@ -75,30 +60,14 @@ static void queue_continuation(struct continuation_info *info) {
   }
 
   else {
-    ready_continuations[(rc_read_ptr + rc_count) % rc_buffer_length] = info;
+    memcpy(
+      &ready_continuations[(rc_read_ptr + rc_count) % rc_buffer_length],
+      info, sizeof(struct continuation_info));
     ++rc_count;
   }
 
 }
 
-//defined in scheduler.asm
-void user_task_start();
-
-void create_user_task(
-  uint64_t cr3, uint64_t rip, uint64_t rsp) {
-
-  struct continuation_info *info =
-    heap_alloc(sizeof(struct continuation_info));
-
-  info->rip = (uint64_t)&user_task_start;
-  info->rsp = (uint64_t)rsp;
-  info->rbx = cr3;
-  info->rbp = rip;
-
-  queue_continuation(info);
-
-}
-
 //defined in scheduler.asm
 [[noreturn]] void resume_continuation(struct continuation_info *info);
 
@@ -107,12 +76,12 @@ void create_user_task(
   while (rc_count == 0)
     __asm__ ("hlt");
 
-  struct continuation_info *info = ready_continuations[rc_read_ptr];
+  struct continuation_info info;
+  memcpy(&info, &ready_continuations[rc_read_ptr], sizeof(struct continuation_info));
 
-  ready_continuations[rc_read_ptr] = 0;
   rc_read_ptr = (rc_read_ptr + 1) % rc_buffer_length;
   --rc_count;
 
-  resume_continuation(info);
+  resume_continuation(&info);
 
 }

src/kernel/scheduler.h

@@ -19,9 +19,20 @@
 
 #include <stdint.h>
 
+struct continuation_info {
+  uint64_t rip;
+  uint64_t rbx;
+  uint64_t rbp;
+  uint64_t rsp;
+  uint64_t r12;
+  uint64_t r13;
+  uint64_t r14;
+  uint64_t r15;
+};
+
 void init_scheduler();
 
-void create_user_task(
-  uint64_t cr3, uint64_t rip, uint64_t rsp);
-
 [[noreturn]] void resume_next_continuation();
+
+//ci is copied
+void add_ready_continuation(struct continuation_info *ci);